Kungfu

Kungfu is an open-source agent infrastructure platform that enables work continuity across different AI agents and execution environments.

Kungfu is an open-source platform designed to ensure work continuity for AI agents, allowing tasks to persist even when switching between different models. It functions by decoupling the work session from the agent, enabling a single workflow to move across different execution surfaces without manual context copying or re-explaining decisions. The system allows users to rotate agents, using different models for varying complexity while keeping the project state intact. It supports failure retention, where interrupted processes remain as attempts within the same project. The architecture includes a mechanism for independent review, separating agent output from the approval process. It is intended for developers who want to own their agent hubs while using Kungfu for underlying state management.

Kungfu at a glance

Best for: Software engineer
Runs on: open source
Pricing: Free
Released: November 2017
Category: AI agent tool
Website: kungfu.tech

Key features

Durable agent work across model changes
Failure retention for crashed processes
Independent review and settlement workflows
Portability across independently owned hubs
Open-source core architecture
Local-first and journal-first design

Use it for

Maintaining workflow state across agent swaps
Reviewing candidate work produced by AI agents
Retaining work progress after connection failures
Switching between different LLMs for specific tasks

Alternatives to Kungfu

Composio — Equips AI agents with tools, managed user auth, and sandboxes across 1,500+ apps
Beeuni — Agents retain context and improve skills over time through a controlled review process.
Rakazo — Complete ownership of data, models, and execution environment.
OpenViking — Unifies agent memory, knowledge RAG, and skills into a single directory structure.
OmniAgent — It evolves skills in real-time and dynamically hardens security during task execution.
OpenSquilla — Maximizes AI intelligence density while minimizing token costs through smart routing.
Goose — Runs locally on your machine with support for 70+ extensions and any LLM.
Raven — Provides a persistent, self-evolving environment for all-domain agent collaboration.
Julep — Builds durable AI agents that maintain state and recover from errors.
FastGPT — It uses hybrid retrieval and auto-cleansed data to prevent hallucinations.
Agent Squad — Runs multi-agent orchestration in Python, TypeScript, and locally on Apple devices.
MyAgent — It features built-in permission controls, long-term memory, and skill self-evolution.

All Kungfu alternatives